Accueil > Actualités > Séminaires > Séminaire de Julien Brajard


Titre : Machine Learning and Ocean Dynamic
Nom du conférencier : Julien Brajard
Son affiliation : LOCEAN
Laboratoire organisateur : LOCEAN
Date et heure : 28-02-2017 11h00
Lieu : UPMC - 4 place Jussieu - 75005 Paris - LOCEAN, tour 45/55, 4eme étage - salle de réunion
Résumé :

The objective of this seminar is to introduce a machine learning approach (also refered as "Big Data" or "Deep learning") to perform a short-term prediction of ocean surface suspended matter without using a numerical model, but only a data-driven model calibrated from a large dataset.The following abstract was presented in the machine-learning section of the AMS annual meeting in January 2017:In the satellite age, geoscientist have acquired an unprecedented aboundance of data describing the earth (ocean and land) surface. This accumulation of observations with high spatio-temporal sampling hasgenerated a demand in ways to optimally extract from these data theuseful features which have the ability to forecast the evolution of some key parameter. In this work we explore the high potential of using advanced machine learning techniques for the prediction of the temporal evolution of 2D oceanographic parameters. We chose to present an experiment on the prediction of sea-surfacefields of the total suspended particulate mater in the english chanell. This choice was motivated by the complexity of the phenomenons impacting this oceanic variable: it is driven both by the neap-tide cycle, storms and general circulation oceanic currents.The predicting system used is constructed using three successive blocks. The first is consisting in a convolutional neural network toextract useful feature and reduce the dimension of the input. The second is a recurrent neural network which is used as a feature predictor. The last block is a convolutional neural network used to reconstruct the image from the predicted feature of the last block. An additional motivation was the frequent missing values caused by the cloud cover over the area. A number of neuronal methods are able to produce good predictions despite missing values. The methodology we selected to implement is a combination of convolutionary neuronal networks and long short-term memory networks. Preliminary results indicate a predictive power for the mean situation and also for extreme events (e.g. storms) than is comparable or better than traditional approaches.

Contact :